Abstract: Present optimization approaches for the mechanical structures usually neglect the uncertainty of material characteristics and the influence of their dynamic characteristics on the overall performance. Moreover, the algorithms for solving the optimization problems are often inefficient. To overcome the above shortcomings, the robust optimization method of the dynamic characteristics of mechanical structures based on double renewal Kriging models is proposed. A robust optimization model including both the means and variances of the mechanical structures’ dynamic characteristics in the objectives is established. The model includes the deformation of mechanical structures as constraints. The kriging models that can accurately and efficiently obtain the values of objective and constraint functions are constructed based on the Latin hypercube experimental design, double update strategy, and the sample point encryption technology in both the local area of maximum error and the neighborhood of the approximate optimal solution. And then all the Pareto optimal solutions to the robust optimization problem of the mechanical structures’ dynamic characteristics are located by integrating a neighborhood cultivation-based genetic algorithm with the double renewal Kriging model. The robust optimal design of the dynamic characteristics of the cone-shape fixed structure of a large turbogenerator’s stator end windings verifies the validity of the proposed method as well as its superiority over the deterministic ones.

Key words: genetic algorithm, mechanical structure;dynamic characteristics;robust optimization;double renewal Kriging model;multi-objective